The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands via Bash to perform git operations (branching, worktrees), create directories, and calculate file checksums. Notably, it uses git worktree add to create new directories in the parent directory (../) of the current project when the --parallel flag is used.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from multiple repository files, including .ai-factory/RESEARCH.md, .ai-factory/DESCRIPTION.md, and project-specific rule overrides in .ai-factory/skill-context/aif-plan/SKILL.md. These inputs directly influence the generated implementation plan and tasks, creating a vector where content in the repository could manipulate the behavior of subsequent agents that execute the plan.
Ingestion points: .ai-factory/config.yaml, DESCRIPTION.md, ARCHITECTURE.md, ROADMAP.md, RESEARCH.md, and skill-context/aif-plan/SKILL.md.
Boundary markers: The skill uses specific markdown sections like ## Original Request and ## Research Context to isolate some inputs, but the skill-context overrides are designed to modify the agent's core instructional logic.
Capability inventory: The skill can write files, execute git commands, manage worktrees, and create new tasks via TaskCreate, which directly drives future agent actions.
Sanitization: The skill performs minimal sanitization, generally copying or interpolating content from the source files into the generated artifacts.
[DYNAMIC_EXECUTION]: The skill utilizes the Task tool to spawn sub-agents for codebase exploration, using dynamically generated prompts based on the user's request and project context.

aif-plan